https://github.com/ayusrjn/ensemble-customer-analyzer

Advanced RoBerta and DistillBert Based Abstract Based Sentiment Analyzer. Ensemble Architecture
https://github.com/ayusrjn/ensemble-customer-analyzer

distillbert-model ensemble roberta-base setfit-absa

Last synced: 3 months ago
JSON representation

Advanced RoBerta and DistillBert Based Abstract Based Sentiment Analyzer. Ensemble Architecture

• Host: GitHub
• URL: https://github.com/ayusrjn/ensemble-customer-analyzer
• Owner: ayusrjn
• Created: 2025-01-31T18:03:15.000Z (over 1 year ago)
• Default Branch: main
• Last Pushed: 2025-01-31T19:16:24.000Z (over 1 year ago)
• Last Synced: 2025-01-31T19:28:45.308Z (over 1 year ago)
• Topics: distillbert-model, ensemble, roberta-base, setfit-absa
• Language: Python
• Homepage:
• Size: 0 Bytes
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

          
# RoBerta and DistillBert Based Advanced Abstract Based Sentiment Analysis App

## Overview

An advanced sentiment analysis system trained on Flipkart product reviews, combining the power of RoBERTa and DistilBERT models in an ensemble approach. The system not only classifies sentiment but also extracts key aspects from negative reviews for better insights.

## Application Screenshots

![Main Dashboard](assets/dashboard.png)

*Main application dashboard showing sentiment distribution*

![Results Page](assets/results.png)

*Detailed analysis results with sentiment breakdown*

![Abstract Based Results](assets/absa.png)

*Extracts the context of the reviews*

![Confidence](assets/upload.png)

*gives confidence of the model*

## Features

- **Ensemble Model Architecture**

    - Fine-tuned RoBERTa

    - Fine-tuned DistilBERT

    - Enhanced accuracy through model combination

- **Sentiment Analysis Capabilities**

    - Positive/Negative classification

    - Confidence scores

    - Abstract extraction from negative reviews

- **Review Analysis Features**

    - Aspect-based sentiment extraction

    - Key pain points identification

    - Review summarization

## Usage

```python

# Import required libraries

from transformers import pipeline, AutoTokenizer, AutoModelForSequenceClassification

from sentiment_analyzer import FlipkartAnalyzer

# Load models from Hugging Face Hub

roberta_model = AutoModelForSequenceClassification.from_pretrained("ayusrjn/RoBerta-finetuned-flipkar-review")

roberta_tokenizer = AutoTokenizer.from_pretrained("roberta-base")

distilbert_model = AutoModelForSequenceClassification.from_pretrained("ayusrjn/distillBert-finetuned-on-flipkart-reviews")

distilbert_tokenizer = AutoTokenizer.from_pretrained("distilbert-base-uncased")

```

# Change the model path in the application 

  pages/1_Results.py

  ```python

  // line 66 to 71

  results_df = predict_from_csv(

                csv_path=temp_csv_path,

                text_column='Comments',

                roberta_path=roberta_model,

                distil_path=distilbert,

                num_labels=3)

 ```

# Change absa model path 

absa.py 

  ```python

  model = AbsaModel.from_pretrained(

    "models\setfit-absa-model-aspect",

    "models\setfit-absa-model-polarity"

)

# Download the model from Hugging Face "ayusrjn/setfit-absa/setfit-absa-model-aspect" and "ayusrjn/setfit-absa/setfit-absa-model-aspect"

 ```

## Example File

example.csv

## Installation

```bash

pip install -r requirements.txt

```

## Running

```bash

streamlit run app1.py

```

## License

MIT License